@FGuzmanAI: 56,000+ tokens/sec at just 80 MHz. I burned a full Transformer with KV cache into a custom chip. Designed gate by gate …
Summary
A custom digital chip designed gate-by-gate achieves over 56,000 tokens/sec running a Transformer with KV cache at just 80 MHz, prototyped on an FPGA.
View Cached Full Text
Cached at: 06/13/26, 08:16 PM
56,000+ tokens/sec at just 80 MHz. 🤯
I burned a full Transformer with KV cache into a custom chip. Designed gate by gate as a 100% digital integrated circuit. Prototyped on a FPGA. (No GPU. No CPU) Just pure digital silicon running @karpathy microGPT, spelling out names on a https://t.co/hXX8kKIxiA
Similar Articles
Can Agents Design Better Chips with a Higher Level Abstraction?
This paper explores using LLM agents for chip design with higher-level abstractions, introducing a workflow called AHRR that combines agent-based HLS design with RTL refinement, achieving a 2.6× speedup over direct RTL design in benchmarks.
@TheAhmadOsman: Inference Engineering simply is - Encodings that can be loaded - Operations on said encodings implemented by backends -…
A tweet explains the core components of inference engineering: loadable encodings, backend operations, and native hardware arithmetic, with a discussion on kernels and GPU efficiency.
This cartridge-playing Game Boy clone is smaller and cheaper than Analogue’s Pocket
FunnyPlaying has announced the FPGB Mini, a compact and affordable handheld console that plays Game Boy and Game Boy Color cartridges using FPGA technology, offering a cheaper alternative to the Analogue Pocket.
FairCompressAgent: An Agentic Framework for Fairness-Aware Model Compression for FPGA Deployment
This paper proposes FairCompressAgent, an agentic framework that uses a language-model planner to optimize fairness-aware model compression for FPGA deployment, balancing accuracy, fairness, and efficiency.
Recreating Voodoo Graphics and a Late-1990s Gaming PC on an FPGA
The article details the creation of zSST, a SystemVerilog implementation of 3dfx Voodoo Graphics for FPGA, enabling a DOS PC with Voodoo graphics to run classic games like Tomb Raider.